A DCL3 dicing code within Pol IV-RDR2 transcripts diversifies the siRNA pool guiding RNA-directed DNA methylation

Read the full article See related articles


In plants, selfish genetic elements including retrotransposons and DNA viruses are transcriptionally silenced by RNA-directed DNA methylation. Guiding the process are short interfering RNAs (siRNAs) cut by DICER-LIKE 3 (DCL3) from double-stranded precursors of ∼30 bp synthesized by NUCLEAR RNA POLYMERASE IV (Pol IV) and RNA-DEPENDENT RNA POLYMERASE 2 (RDR2). We show that Pol IV initiating nucleotide choice, RDR2 initiation 1-2 nt internal to Pol IV transcript ends and RDR2 terminal transferase activity collectively yield a code that influences which end of the precursor is diced and whether 24 or 23 nt siRNAs are generated from the Pol IV or RDR2-transcribed strands. By diversifying the size, sequence, and strand polarity of siRNAs derived from a given precursor, alternative patterns of DCL3 dicing allow maximal siRNA coverage at methylated target loci.

Article activity feed

  1. Evaluation Summary:

    The paper is of interest to RNA biologists, especially to those who study small RNAs. The findings deepen our understanding of the rules of DCL3 dicing and explain how 23-nt and 24-nt siRNAs in the RNA-directed DNA methylation pathway are produced. Overall, the data are of high quality and support the paper's conclusions.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #3 agreed to share their name with the authors.)

  2. Reviewer #1 (Public Review):

    The authors use an array of in vitro experiments to evaluate the rules that govern 23/24-nt sRNA production by DCL3. The authors conclude that DCL3 measures from only one end of a dsRNA precursor, prefers ends with overhangs and terminal A or U nucleotides, and does not require ATP. The authors also conclude that RDR2 generates overhangs not only via its terminal transferase activity, but by initiating synthesis 1 or 2 nucleotides from the end of the molecule synthesized by Pol IV. This enables 24-nt production from both the Pol IV and RDR2 strands of the dsRNA template.

    The paper's conclusions are generally well-supported.

  3. Reviewer #2 (Public Review):

    Loffer et al. investigated the activity of recombinant DCL3 against substrates with overhangs or blunt ends and substrates with different 5' nucleotide. They found that DCL3 preferentially cuts dsRNAs with 3' overhangs and dsRNAs with 5' AU pair and monophosphate group. DCL3 measures 24 nt from the 5' end of one strand to make the first cut and chooses a position 2 nt offset on the opposite strand to make the second cut. Investigation of the activity of mutant forms of DCL3 revealed that RNase III domain B of DCL3 cuts the measured strand, while domain A cuts the non-measured strand. Interestingly, they found that 3' overhangs are produced at both ends of dsRNAs because RDR2 initiates transcription internally and also has terminal transferase activity and 23-nt siRNAs and 24-nt siRNAs can be produced both from the Pol IV strand and the RDR2 strand. It should be noted that end point assays were used for dicing assays in Fig. 4A-C (three time points) and Fig. 5 (one time point). The results from the end point assays might be misleading, especially when the substrates were not excessive (Fig. 4C and Fig. 5). Extra care needs to be taken to draw the conclusion that ATP is not required for DCL3 activity. In the discussion, the authors proposed that Pol IV, RDR2 and DCL3 activities can collectively account for much of the 5' A-bias among siRNAs that become loaded into AGO4, independent of AGO4-mediated selection. A better interpretation could be that the initiation (production of 5' A siRNAs) and effector (specific binding to AGO4 that preferably accepts 5' A siRNAs) stages of RdDM have co-evolved to achieve the specificity of RdDM pathway.

  4. Reviewer #3 (Public Review):

    RdDM is the most important mechanism for understanding plant epigenetics. In order to maintain the plant genome integrity, heterochromatin regions containing transposons and near centromeres are essentially stabilized by RdDM. Although a large number of genes (proteins) involved in RdDM have been genetically identified, the functions of these proteins have not been characterized in detail. In this paper, biochemical analyses using purified recombinant DCL3 show that DCL3 efficiently cleaves short dsRNAs with 1- to 2-nt 3' overhang (OH) and the 5' phosphate. The authors also show that DCL3 preferentially cleaves dsRNAs with adenosine (A) or uridine (U) at the 5' phosphate end. And they show that RNA-dependent RNA polymerase 2 (RDR2) initiates second strand RNA synthesis 1- to 2-nt internal to the RNA polymerase IV (PolIV) transcript and has terminal transferase activity to add one nucleotide at the 3' end of its transcripts. These dsRNAs with the 3'overhangs at both ends are preferred DCL3 substrates. Finally, they propose that these RDR2 and DCL3 activities shown here together with the results of their previous study (Pol IV initiating nucleotide choice, Blevins et al., 2015; Singh et al., 2019) can collectively account for much of the 5' A-bias among siRNAs that become loaded into argonaute 4 (AGO4).

    The Pikaard's group has achieved excellent research by analyzing enzymatic activities of proteins involved in RdDM such as PolIV, RDR2 and DCL3, though it is very difficult to purify these high-molecular-weight proteins to analyze their enzymatic activities. In this paper, the excellent biochemical results are shown by using purified recombinant proteins and polyacrylamide gel electrophoresis (PAGE) with high resolution.